This study compared the accuracy of physical activity energy expenditure (PAEE) prediction using two methods of accounting for age dependency versus one standard (single) value across all ages.

PAEE estimates were derived by pooling data from five studies. Participants, 6–18 years (n=929), engaged in 14 activities while in a room calorimeter or wearing a portable metabolic analyzer. Linear regression was used to estimate the measurement error in PAEE (expressed as MET_{y}) associated with using age-groups (6–9, 10–12, 13–15, and 16–18 years) and age-in-years (each year of chronological age (e.g., 12=12.0–12.99 years)) versus the standard (a single value across all ages).

Age-groups and age-in-years showed similar error, and both showed less error than the standard method for cycling, skilled and moderate-to-vigorous intensity activities. For sedentary and light activities, the standard had similar error to the other two methods. Mean values for root mean square error ranged from 0.2–1.7 MET_{y} across all activities. Error reduction ranged from −0.2–21.7% for age-groups and −0.23–18.2% for age-in-years, compared to the standard.

Accounting for age showed lower errors than a standard (single) value; using an age-dependent model in the Youth Compendium is recommended.

The Compendium of Physical Activities standardized the coding of physical activity energy expenditure (PAEE) by activity type and intensity (METs) for adults (

Despite the fact that evidence has shown adult MET values both underestimate and overestimate PAEE for youth, depending on activity (

To update the Youth Compendium with youth-based values, a method is needed that expresses the energy cost of physical activities and considers: 1.) children’s resting energy expenditure is higher than adults’ and declines with age on a mass-specific basis (_{2}/min) but decreases with age when expressed relative to body mass (mlO_{2}/kg/min) (

Previously, our work group examined expressions of energy cost and suggested that the Youth MET (MET_{y}; energy cost of the activity divided by (estimated) basal energy expenditure) was the most practical metric to use in an updated Youth Compendium (_{y} does not completely ameliorate age-related changes in energy expenditure. Thus, given the wide range of energy equivalences for one MET over childhood and adolescence, choosing one standard MET value to apply to all ages may be challenging. Alternatively, employing several values to represent MET_{y} across the age spectrum may be unwieldy in practice. The method of expressing the energy cost of physical activities should provide the smallest error possible while considering age-related differences in energy cost in children and adolescents. Therefore, the purpose of this study was to compare the measurement error for two methods of accounting for age dependency in MET_{y}, age-groups (6–9, 10–12, 13–15, and 16–18 years) and age-in-years (e.g., 12=12.0–12.99), to a standard for all ages. This is necessary because the development of an updated Youth Compendium is underway.

Investigators from five sites contributed the data included in this manuscript: Baylor College of Medicine (_{2} measurement system used also differed by study site: MSU and OSU (Oxycon portable metabolic analyzer), UMB and UNC (Cosmed portable metabolic analyzer), and Baylor (room calorimeter). All systems were calibrated with tanks containing known concentrations of O_{2} and CO_{2}, and flow rates were calibrated with a 3-L syringe in the case of portable metabolic systems or infusion of N_{2} and CO_{2} into the room calorimeters to simulate human respiration. Data collected with portable metabolic analyzers were collected breath-by-breath and averaged over 1-min time intervals. Data collected with the room calorimeter were also averaged over 1-min time intervals.

In the combined data set, there were 43 activities with measured values. Fourteen activities with ≥ 100 participants and an age range of at least 6 to 18 years were selected for analysis. The 14 activities included were computer games, television viewing, housework, sweeping, Wii play, cycling, aerobics, dance, walking at 2 and 3 mph, running at 4 and 5 mph, basketball, and rope skipping. Specific details regarding exactly how activities were performed can be found in the original manuscripts from each site (

Across all studies, participants ranged from 5 to 19 years (overall mean 12 ± 3 y (mean ± SD)). The total sample included 1060 youth. However, 112 individuals did not have VO_{2} data for any of the 14 activities or were missing key variables such as age, height, or body mass, and were excluded from analyses. Additionally, the research team excluded data from 19-year-olds (n=2) and five-year-olds (n=17). The final sample included 505 boys and 424 girls, for a total of 929 youth. Data from MSU and OSU included participants 6–15 years old. UNC included 8–18 year-olds. UMB included 11–15 year-olds, and Baylor included 6–18 year-olds. All activities included in analyses involved participants 6–18 years old, with the exception of basketball (7–16 years) and rope skipping (8–18 years). See

Variables included demographic characteristics (age, sex), anthropometric characteristics (height, body mass), and (absolute) VO_{2} values (in ml∙min^{−1}) for all activities. Height and body mass were used to calculate body mass index (BMI) and BMI percentile according to reference values from the Centers for Disease Control and Prevention (CDC) growth charts (

The Youth MET (MET_{y}) was calculated for each participant for each activity as EE (kcal∙kg^{−1}∙hr^{−1}) divided by basal metabolic rate (BMR), which was determined using the Schofield equation (

We compared the error associated with the age-groups and age-in-years methods to a method with a standard value across ages (which was the referent group; _{y} value, representing the sample’s mean age, was assigned to a specific activity, regardless of age.

Standard (metric constant): Metric = b0

For the age-groups method, a MET_{y} value was assigned to each of the four age-groups using an average age for each age group: 6–9 year olds, 10–12 year olds, 13–15 year olds, and 16–18 year olds. The four age-groups were chosen to represent pre- (middle childhood), early, mid- and late adolescence (

Age-groups (metric standard within age-groups): Metric = b0 + b1*10–12 years + b2*13–15 years + b3*16–18 years, referent group is 6–9 years

The age-in-years method allowed the metric to vary for each year of age (e.g., 12-yr-olds were 12.0–12.99 years). The predicted value is a standard plus an incremental value for each additional year.

Age-in-years (metric varies with age): Metric = b0 + b1*Age (years)

Following the analyses, the team developed an example layout for the Youth Compendium (

Average height of the sample was 152 ± 17 cm (mean ± SD), and average body mass was 49 ± 20 kg. BMI averaged 21 ± 7 kg∙m^{−2}, with 67% of the sample classified as normal weight (5^{th}-84^{th} percentile), 15% overweight (85^{th}-94^{th} percentile), and 18% obese (≥95^{th} percentile), according to Centers for Disease Control and Prevention growth charts (_{2} values) and only removed those values that were deemed biologically implausible (which in all cases were more than two standard deviations from the age- and sex-specific mean for that activity). Basketball was the only activity for which there was a significant (p˃0.01) site by age interaction. Upon further investigation, it appeared the difference was due to differences in sample site characteristics. One site had an equal number of boys and girls whereas the other site had twice as many boys as compared to girls. An additional examination involved fitting the basketball activity models with a site main effect and a site by age interaction. The results yielded even smaller error, similar to the model without site, with <1% difference in the percent reduction of error between the two models accounting for age.

There was a greater reduction in error for MET_{y} using age-groups or age-in-years for activities such as cycling and more vigorous intensity/skilled activities (_{y}) for sedentary and light activities. Error for cycling, moderate-to-vigorous, and activities requiring complex motor skills (e.g., basketball) was higher (0.6 to 1.7 MET_{y}). Additionally, for all cycling, moderate-to-vigorous, and skilled activities (with the exception of walking at 3 mph), age-groups and age-in-years methods had lower error relative to the standard method.

_{y} in older children compared to younger children for basketball than for walking.

_{y} between the standard and age-groups methods. Using a standard MET_{y} value for all ages (6–18 y) results in overestimation of EE for younger age-groups and underestimation of EE for older age-groups for most activities. For some activities, such as sedentary and light, differences across age-groups were less than one MET_{y}. However, for skilled activities, differences across age-groups could be as much as 2–3 MET_{y}, illustrating that the use of MET_{y} by age-groups would result in less over- and under-estimation than the use of a standard MET_{y} value.

Our findings showed that using MET_{y} values that accounted for age resulted in less error in energy expenditure prediction than using one standard value across all ages, particularly for cycling, moderate-to-vigorous, and skilled activities. Thus, a single value, like in the Adult Compendium (

The original Ridley Compendium provides a single, standard MET_{y} value for all ages (_{y} values from the current study to the Ridley Compendium, the values appear to be similar for sedentary and light activities, but some values are higher and some are lower for locomotor activities and bicycle riding. Regardless, the approach of using one value across ages is problematic given the known energy expenditure differences at rest and during activity across ages. For example, an 8-year-old child has a BMR of approximately 1.7 kcal.kg^{−1.}h^{−1} while a 16-year-old adolescent has a BMR of approximately 1.1 kcal.kg^{−1.}h^{−1} (_{y} would have energy expenditures of 10.2 kcal.kg^{−1.}h^{−1} and 6.6 kcal.kg^{−1.}h^{−1} for the 8- and 16-year-old, respectively, which is a difference of 3.6 kcal.kg^{−1.}h^{−1} or approximately 35% less for the 16 year old. If one assumes the body mass of a 10-year-old is 35 kg then the child would expend ~357 kcal for an hour of bicycle riding, whereas a 60 kg, 16-year-old would expend ~396 kcal doing the same activity. This amounts to approximately an 11% higher PAEE for the 16-year-old adolescent compared to the 8-year-old child. Additionally, younger children are less efficient when performing many activities because they have higher EE per unit mass compared to an adolescent (_{y} in the 6-yr-old vs. 16-yr-old, respectively) in activities such as cycling. Using the methods that take age into account reduces error up to 15–20% compared to the standard. Thus, we do not recommend a single, standard MET_{y} value for activities included in the Compendium.

The age-groups and age-in-years methods provide similar energy expenditure estimates for most activities. Since BMR changes significantly as the child grows, a specific age-related methodology would be optimal, especially for research purposes. However, the age-in-years approach may be difficult for some users of the Compendium, such as practitioners in an applied setting. The age group approach has been used previously by Harrell et al. (

This study had several strengths and weaknesses. One weakness was a small number of younger or older children for some activities. Rope skipping and cycling included fewer 6–9 year-olds than other age-groups, while housework and basketball included fewer 16–18 year-olds. This may have resulted in less variability in the results because of more similar BMR and metabolic efficiency in the remaining age-groups. Another weakness was use of the Schofield equation to estimate BMR, rather than using measured values. Although the different data collection sites collected pre-exercise (resting) data, the protocols varied and all measures were taken immediately pre-exercise. True measures of BMR are difficult to obtain in children, and although it would be preferable to have them the team considered it better to standardize based on the equation. Similarly, most studies that would need to use a Youth Compendium would not be likely to assess true BMR.

We did not present values separately by sex. The Adult Compendium does not distinguish between sexes, either, and when sex differences were examined in the current sample, they comprised less than two percent of the variance in energy expenditure values after correcting for age and body mass. Also, some may view lack of accounting for body composition as a limitation. However, the adult Compendium does not make a weight/body composition-based adjustment. The purpose is to apply on a population-level basis, meaning that many types of individual differences are not accounted for in a Compendium. However, if researchers or practitioners desire to use an individualized approach, a corrected MET adjustment procedure is available for the adult Compendium (

Use of age-groups or age-in-years MET_{y} values produces estimation errors that are similar and acceptable. However, assigning a standard MET_{y} value to represent energy cost of a given activity for all ages of children and adolescents is not optimal. Using age-groups to express energy expenditure in an updated youth Compendium may be a feasible alternative that also provides acceptable accuracy.

The authors wish to thank Dr. Russell R. Pate for his assistance in developing the context and design of the study and Anne B. Rodgers for editorial considerations.

No funding was received for this project. The authors gratefully acknowledge the National Collaborative on Childhood Obesity Research (NCCOR) and the staff of FHI360 for administrative and logistical support.

The findings and conclusions in this article are those of the authors and do not necessarily represent the official positions of the CDC or NCCOR.

The authors report no conflicts of interest.

This article will be published in a forthcoming issue of the

Visual representation of the three models used to examine the additional reduction in error in Youth METs (MET_{y}) for walking 2 mph when accounting for age: standard MET_{y} across all ages (top), standard MET_{y} for age-groups (middle), and age-in-years MET_{y} (bottom).

Visual representation of the three models used to examine the additional reduction in error in Youth METs (MET_{y}) for basketball when accounting for age: standard MET_{y} across all ages (top), standard MET_{y} for age-groups (middle), and age-in-years MET_{y} (bottom). Note data were not collected for 17–18 year olds.

Number of participants included in activities by age group

Activity | 6–9 | 10–12 | 13–15 | 16–18 |
---|---|---|---|---|

| ||||

Computer games | 102 | 136 | 95 | 30 |

Television viewing | 101 | 125 | 110 | 89 |

| ||||

Housework | 68 | 99 | 61 | 5 |

Sweeping | 124 | 170 | 133 | 74 |

Wii® Play | 44 | 58 | 49 | 20 |

| ||||

Cycling ~10 mph | 59 | 67 | 75 | 75 |

| ||||

Aerobics | 114 | 118 | 92 | 21 |

Dance | 45 | 62 | 48 | 25 |

Walk - 2 mph | 132 | 132 | 138 | 107 |

Walk - 3 mph | 120 | 47 | 7 | 98 |

Run - 4 mph | 78 | 83 | 84 | 70 |

Run - 5 mph | 44 | 29 | 12 | 23 |

| ||||

Basketball | 105 | 89 | 61 | 8 |

Rope Skipping | 49 | 77 | 72 | 65 |

The number of participants (and percentage) for demographic characteristics by age group

Participant | 6–9 years | 10–12 years | 13–15 years | 16–18 years | 6–18 years | |||||
---|---|---|---|---|---|---|---|---|---|---|

Characteristics | (n=241) | (n=329) | (n=242) | (n=117) | (n=292) | |||||

Sex | ||||||||||

Boys | 124 | (51.5) | 178 | (54.1) | 147 | (60.7) | 56 | (47.9) | 505 | (54.4) |

Girls | 117 | (48.5) | 151 | (45.9) | 95 | (39.3) | 61 | (52.1) | 424 | (45.6) |

Race/Ethnicity | ||||||||||

African American | 31 | (12.9) | 90 | (27.4) | 57 | (23.6) | 18 | (15.4) | 196 | (21.1) |

Caucasian | 169 | (70.1) | 175 | (53.2) | 141 | (58.3) | 79 | (67.5) | 564 | (60.7) |

Hispanic | 30 | (12.4) | 38 | (11.6) | 27 | (11.2) | 10 | (8.5) | 105 | (11.3) |

Other | 11 | (4.6) | 26 | (7.9) | 17 | (7.0) | 10 | (8.5) | 64 | (6.9) |

Body Mass Index Category | ||||||||||

Underweight | 5 | (2.1) | 8 | (2.4) | 3 | (1.2) | 1 | (0.9) | 17 | (1.8) |

Normal Weight | 170 | (70.5) | 189 | (57.4) | 157 | (64.9) | 84 | (71.8) | 600 | (64.6) |

Overweight | 28 | (11.6) | 64 | (19.5) | 34 | (14.0) | 17 | (14.5) | 143 | (15.4) |

Obese | 38 | (15.8) | 68 | (20.7) | 48 | (19.8) | 15 | (12.8) | 169 | (18.2) |

Body mass index for age is based on measured height and weight. Underweight = <5th percentile, healthy weight = 5 - <85^{th} percentile; overweight ≥85^{th} - < 95^{th} percentile; obese = ≥95^{th} percentile (

Note: All column percentages may not add up to 100% because of rounding

Results from comparison of estimated youth METs from standard, age-groups, and age-in-years models

Activity | N | Model Descriptive Statistics | Model Root Mean Square Error (RMSE) | Percent Reduction in RMSE from Standard | ||||||
---|---|---|---|---|---|---|---|---|---|---|

All Models | Standard | Age-Groups | Age-in-Years | Standard | Age-Groups | Age-in-Years | Age-Groups | Age-in-Years | ||

| ||||||||||

Mean | (SD) | (SD) | (SD) | RMSE (RSD) | RMSE (RSD) | RMSE (RSD) | % | % | ||

| ||||||||||

Computer games | 363 | 1.40 | (0.23) | (0.01) | (0.03) | 0.23 (0.31) | 0.23 (0.31) | 0.23 (0.32) | 1.59 | 0.56 |

Television viewing | 425 | 1.20 | (0.26) | (0.05) | (0.03) | 0.26 (0.34) | 0.25 (0.33) | 0.25 (0.33) | 1.76 | 0.65 |

| ||||||||||

Housework | 233 | 3.00 | (0.57) | (0.04) | (0.05) | 0.57 (0.83) | 0.57 (0.81) | 0.57 (0.81) | −0.22 | 0.22 |

Sweeping | 501 | 3.32 | (0.63) | (0.06) | (0.08) | 0.63 (0.79) | 0.62 (0.80) | 0.62 (0.80) | 1.22 | 0.82 |

Wii® Play | 171 | 2.63 | (0.72) | (0.11) | (0.03) | 0.72 (1.08) | 0.70 (1.03) | 0.72 (1.08) | 1.37 | −0.23 |

| ||||||||||

Cycling ~10 mph | 276 | 5.76 | (1.30) | (0.43) | (0.65) | 1.29 (1.51) | 1.09 (1.31) | 1.12 (1.29) | 15.26 | 13.52 |

| ||||||||||

Aerobics | 345 | 3.83 | (0.98) | (0.21) | (0.36) | 0.98 (1.12) | 0.92 (1.06) | 0.92 (1.07) | 6.16 | 6.65 |

Dance | 180 | 3.55 | (0.98) | (0.16) | (0.25) | 0.97 (1.23) | 0.91 (1.12) | 0.94 (1.17) | 5.68 | 3.11 |

Walk - 2 mph | 509 | 3.56 | (0.66) | (0.23) | (0.17) | 0.66 (0.84) | 0.61 (0.77) | 0.64 (0.80) | 7.25 | 3.08 |

Walk - 3 mph | 272 | 4.41 | (0.92) | (0.46) | (0.57) | 0.92 (1.66) | 0.92 (1.62) | 0.92 (1.64) | 0.67 | 0.41 |

Run - 4 mph | 315 | 8.16 | (1.56) | (0.62) | (0.90) | 1.56 (1.92) | 1.22 (1.47) | 1.27 (1.61) | 21.66 | 18.25 |

Run - 5 mph | 108 | 8.21 | (1.61) | (0.11) | (0.08) | 1.60 (2.43) | 1.53 (2.26) | 1.50 (2.21) | 4.26 | 6.51 |

| ||||||||||

Basketball | 263 | 6.77 | (1.63) | (0.71) | (0.91) | 1.62 (1.85) | 1.35 (1.55) | 1.35 (1.58) | 16.21 | 16.95 |

Rope Skipping | 263 | 8.10 | (1.66) | (0.18) | 0.64) | 1.66 (2.04) | 1.50 (1.87) | 1.53 (1.85) | 8.71 | 7.60 |

Standard Deviation (SD) from sample, SD for model with constant is zero

Note: RMSE is the root of the mean square errors; percent reduction reflects the relative reduction in the root mean square error (RMSE) from the model that is constant across all ages compared to the models that include age. For example, percent reduction from Standard to Age-in-years is (RMSE_{constant}- RMSE_{ageyrs})/RMSE_{constant} x 100%. Positive % indicates a decrease in error and a negative % indicates an increase in error when including age.

Youth MET (MET_{y}) values for ages 6–18 years based on regression analysis

Activity | Standard MET_{y}^{a} | MET_{y} by Age Group^{b} | Age-in-years: Equation (MET_{y}=b0 + b1*age) | |||
---|---|---|---|---|---|---|

6–18 years | 6–9 years | 10–12 years | 13–15 years | 16–18 years | ||

| ||||||

Computer games | 1.40 | 1.44 | 1.41 | 1.38 | 1.35 | 1.51 + −0.01 x age |

Television viewing | 1.20 | 1.16 | 1.19 | 1.22 | 1.25 | 1.08 + 0.01 x age |

| ||||||

Housework | 3.02 | 2.92 | 3.00 | 3.06 | 3.13 | 2.76 + 0.02 x age |

Sweeping | 3.32 | 3.44 | 3.35 | 3.26 | 3.18 | 3.66 + −0.03 x age |

Wii® Play | 2.63 | 2.59 | 2.62 | 2.65 | 2.67 | 2.53 + 0.01 x age |

| ||||||

Cycling ~10 mph | 5.60 | 4.75 | 5.41 | 5.98 | 6.54 | 3.34 + 0.19 x age |

| ||||||

Aerobics | 3.95 | 3.38 | 3.82 | 4.20 | 4.58 | 2.44 + 0.13 x age |

Dance | 3.56 | 3.19 | 3.48 | 3.72 | 3.97 | 2.57 + 0.08 x age |

Walk - 2 mph | 3.56 | 3.34 | 3.51 | 3.66 | 3.81 | 2.96 + 0.05 x age |

Walk - 3 mph | 4.40 | 4.27 | 4.37 | 4.46 | 5.55 | 4.04 + 0.03 x age |

Run - 4 mph | 8.10 | 6.89 | 7.83 | 8.64 | 9.45 | 4.86 + 0.27 x age |

Run - 5 mph | 8.27 | 7.33 | 8.06 | 8.69 | 9.32 | 5.75 + 0.21 x age |

| ||||||

Basketball | 7.28 | 5.79 | 6.95 | 7.94 | 8.93 | 3.31 + 0.33 x age |

Rope Skipping | 7.93 | 7.01 | 7.73 | 8.34 | 8.96 | 5.46 + 0.21 x age |

Mid-point of age range is 12 years

Mid-points for age-groups are 7.5 years, 11 years, 14 years, and 17 years